Prevalent and Diverse Intratumoral Oncoprotein-Specific CD8⁺ T Cells within Polyomavirus-Driven Merkel Cell Carcinomas

Merkel cell carcinoma (MCC) is a rare, clinically aggressive cutaneous malignancy. Surgery and radiation can clear early-stage disease, but persons with nodal or distant spread have nondurable responses to cytotoxic chemotherapy and a poor prognosis (1). Studies have changed our understanding of MCC and improved outcomes for some patients. Feng and colleagues discovered that most MCC tumors are associated with Merkel cell polyomavirus (MCPyV; ref. 2). MCPyV infection is ubiquitous in healthy persons (3, 4). Rarely, viral DNA integrates into diverse, apparently random genomic loci, with further T-antigen (T-Ag) mutations promoting oncogenesis (2). Immune checkpoint inhibitors that block ligand interactions with programmed death-1 (PD-1) have led to deep, durable responses for MCPyV⁺ MCC, and this therapy is now standard for advanced disease (5, 6). Local immunotherapy with toll-like receptor agonists, cytokines, oncolytic virus, and systemic infusion of MCPyV-specific CD8⁺ T cells can also mediate responses (7–10). However, many patients show no durable improvement. Thus, an interest exists in using vaccination to augment tumor immunity (11–13).

The MCPyV T gene encodes the nonspliced, 186 amino acid (AA) small T antigen (ST), and longer, spliced large T antigen (LT). The shared C-terminal domain is common T antigen (CT), with the CT/LT splice point at AA 78/79. In tumors, a mutation near LT AA 280-320 truncates LT protein, blocking viral replication (14). ST protein is also expressed in tumors. Together, the T-Ag contains about 400 AAs of unique polypeptide (2) and has high expression in tumors, as measured by IHC. Evidence points to a dominant role for the ST in tumor initiation (15), whereas continuing LT expression is required to maintain cell proliferation in vitro (16). Because T-Ag promotes cell proliferation, DNA, RNA, or vectored vaccines should be modified to reduce oncogenicity risk.

Interestingly, antibodies to T-Ag are undetectable in healthy, MCPyV-infected persons, likely reflecting nuclear localization and low inflammation in areas of replication. In contrast, patients with MCPyV⁺ MCC have high T-Ag–specific IgGs (17) in serum, which are a clinically useful tumor burden biomarker (17). Circulating T-Ag–specific CD8⁺ T cells are similarly detected by ex vivo peripheral blood mononuclear cell (PBMC) tetramer stains in patients with MCPyV⁺ MCC with large tumors and then decrease after successful therapy. This temporal pattern, and phenotypic and functional data for PD-1 expression by MCPyV-specific CD8⁺ T cells (18) are consistent with dysfunction and provide a rationale for immune checkpoint blockade (5).

Taken together, these data and the correlation between CD8⁺ T-cell infiltration/mRNA signatures and clinical outcomes (19, 20) indicate that T-Ag–specific T cells may participate in antitumor responses. Our laboratory and others (20–23) have detected a limited number of MCPyV T-Ag CD8⁺ T-cell epitopes and have harnessed these for adoptive T-cell therapy (7, 24). Monotherapy with HLA-A– or HLA-B–restricted CD8⁺ T cells leads to immune escape due to selective transcriptional downregulation of the relevant HLA (24), suggesting that multi-epitope, multi-HLA therapy might be more effective. Data from active vaccination for HPV-related malignancies and combined tumor vaccination with immune checkpoint blockade (25) suggest that active vaccination is rational for MCPyV⁺ MCC, but it is unknown what proportion of patients have the immunogenetic potential to respond, or how best to modify a genetic vaccine to avoid vaccine-related oncogenesis.

Here, we identified 16 MCPyV CD8⁺ T-cell epitopes, their HLA-restricting alleles, associated tetramer reagents, and T-Ag–specific T-cell receptors (TCR). Together, these HLA alleles covered about 97% of our cohort of patients, indicating that active vaccination may be a rational component of immunotherapy. The clustering of MCPyV CD8⁺ T-cell epitopes in certain regions of the LT suggests that vaccines may be able to eliminate some T-Ag regions while retaining good population coverage.

Fourteen subjects (Supplementary Table S1) with biopsy-confirmed MCC were staged per the American Joint Committee on Cancer Staging Manual 8th Edition (26). Subjects 659 and 861 had biopsies performed elsewhere, which were overnight shipped at ambient temperature in T-cell medium (TCM; ref. 27). Normal tissue biopsies were not studied. TCM is RPMI1640 with HEPES, containing 1% penicillin-streptomycin, 2 mmol/L l-glutamine, 5% fetal calf serum (all Thermo Fisher Scientific) and 5% human serum (HS1004CHI, Valley Biomedical). Tumor-infiltrating lymphocyte (TIL) culture was initiated immediately after biopsy receipt in the laboratory. TILs were generated from biopsies obtained for diagnosis, excision, or in some case at defined time points during clinical trials (Supplementary Table S1). Inclusion criteria for this study were a diagnosis of MCPyV⁺ MCC and successful in vitro expansion of TIL. Exclusion criteria was any prior intralesional immunotherapy, or systemic immune checkpoint blockade prior to the biopsy used for TIL generation. PBMCs were also enriched from peripheral blood for some patients by Ficoll density gradient centrifugation (28) and cryopreserved in LN2 (liquid nitrogen). All subjects gave written informed consent and research was conducted per the principles of the Declaration of Helsinki.

Formalin-fixed, paraffin-embedded tumor sections were diagnosed as MCC by characteristic hematoxylin and eosin (H&E) staining histopathology using an Olympus BX51 microscope (Olympus) and IHC staining for cytokeratin 20 (clone Ks20.8, Dako). In general, H&E stains of MCC exhibit small round blue tumor cells with high nuclear to cytoplasm ratio, round or oval nuclei, finely dispersed chromatin, indistinct nucleoli and scant cytoplasm, conspicuous mitoses and apoptotic bodies, and variable nuclear molding. All subjects had positive serum antibodies to T-Ag (17). These antibodies are very rare in healthy individuals and their presence correlates well with detection of MCPyV in tumors (2). Subject 659 (Supplementary Table S1) had an Allred score of 8 for MCPyV T-Ag protein detection by IHC (19). Staining used mAb CM2B4 (Santa Cruz Biotechnology) at 1 μg/mL. T-Ag stains were not done on the other subjects.

HLA typing was performed at Bloodworks (formerly Puget Sound Blood Center) or by Scisco Genetics, using blood. The performing lab for subjects from whom TILs were studied is indicated in Supplementary Table S1. Bloodworks use PCR followed by sequence-specific oligonucleotide probes, while Scisco Genetics uses PCR followed by next-generation sequencing. For analysis of the extended cohort of 131 patients with MCPyV⁺ MCC, we included HLA data from referring physicians, who used locally available methods. If only two-digit HLA typing data were available, we imputed the presence of the most population prevalent AA-level allelic variant; for example, HLA A*02 was treated as HLA A*02:01. To score each subject for the presence of one or more HLA allelic variant for which MCPyV CD8⁺ T-cell epitopes were defined, HLA data were entered into a spreadsheet (Excel, Microsoft) and the “countif” function used. For MCPyV⁺ MCC cell lines, MKL-1 cells have the genotype HLA A*03:02:01, A:11:01:01; HLA B*08:01:01, B:15:01:01; HLA C*04:01:01, C*07:02:01, and MS-1 cells have the genotype HLA A*02:01:01, A:25:01:01; HLA B*08:01:01, B*15:01:01; HLA C*03:04:01, C*07:01:01, both as determined at Scisco Genetics.

Tumor tissue was minced and fragments placed in culture with TCM (29), 10⁶ allogeneic irradiated PBMC feeder cells, Remel phytohemagglutinin-purified (1.6 μg/mL, PHA-P) mitogen (Thermo Fisher Scientific), human recombinant (hr) IL15 (10 ng/mL; R&D Systems), and human natural IL2 (32 U/mL; Hemagen; ref. 30). At two weeks, bulk TILs were polyclonally reexpanded using PBMCs and EBV-lymphocyte continuous line (LCL) as feeder cells, anti-CD3 mAb as mitogen, and hrIL2 (27). The protocol has been detailed (28). In brief, each 25 cm² cell culture flask contained 2.5 × 10⁷ irradiated allogeneic PBMC, 5 × 10⁶ irradiated allogeneic EBV-LCL, anti-CD3 mAb clone OKT3 (30 ng/mL; Thermo Fisher Scientific), and 10⁵ TIL cells in 25 mL TCM. On day 1, recombinant human IL2 (50 U/mL; aldesleukin, Prometheus Therapeutics, purchased from University of Washington clinical pharmacy) was added. Expansion was continued for 14–16 days followed by cryopreservation in aliquots in LN₂.

We modified methods (28, 31) previously detailed for other viruses. Relevant HLA class I allele cDNAs were cloned into pCDNA3.1 (Thermo Fisher Scientific) using RT-PCR with subject PBMC-derived RNA and HLA locus–specific primers. The sequences of the cDNA clones were checked by Sanger dideoxy sequencing and were identical to allele-specific sequences in IMGT (32). In brief, HLA cDNA were either cloned from PBMC from HLA-typed persons or obtained ready-to-use from the International Histocompatibility Working Group gene bank housed at Fred Hutchinson Cancer Research Center (FHCRC, Seattle, WA; https://www.fredhutch.org/en/research/institutes-networks-ircs/international-histocompatibility-working-group.html). The protocol for cloning of HLA cDNA from PBMC has been previously detailed and was performed exactly as described previously (28). Primers for HLA cDNA PCR are provided in Supplementary Table S2.

MCPyV LT AA 1-327 or full-length ST AA 1-186 with a carboxy-terminus six-histidine addition were cloned into Nature Technology Corporation (NTC) 9385R (ref. 33; Nature Technology Company). These plasmids encode an identical 78 AA N-terminal CT domain. The NTC plasmids were based on Genbank HM011538.1. LT AA 1-259 from MCVw156 (Genbank HM355825.1) was separately cloned into pDEST103, a vector constructed in our laboratory (31), with a CMV promoter that is suitable for transient transfection of eukaryotic cells such as the Cos-7 cells used herein. To create artificial antigen-presenting cells (aAPC), Cos-7 cells (ATCC) were plated at 1.1 × 10⁴ cells/well in 96-well flat-bottom plates (27) and cotransfected for 24 hours with 0.5 μL Fugene 6 (Thermo Fisher Scientific) in 25 μL DMEM (Thermo Fisher Scientific), 75 ng HLA plasmid, and 75 ng MCPyV plasmid/well. Assays were typically duplicate or triplicate with the number of replicates indicated in each figure legend. After two days, 10⁵/well TILs or T-cell clones (TCC), as described below, were added in 200 μL TCM and 24–48 hours later, supernatant IFNγ was measured by ELISA (27). IFNγ results were reported as the mean and the SD of the mean (SD) in the figures.

Epitope mapping initially used, as previously described (34), 95 overlapping T-Ag peptides (OLP; Supplementary Table S3; Genscript) with DMSO stocks stored at −20°C at 2–10 mg/mL. Peptides covering LT AA 1-281 and the unique region of ST were 13 AA long with 9 AA overlap and based on MCPyV 350 (Genbank FJ173805.1; ref. 34). Similar peptides covering LT C-terminal to AA 281, or peptides <13 AA long predicted (35, 36) to bind HLA alleles of interest, were tested in selected cases if we did not observe reactivity to the initial 95 peptides, as detailed in Results section. For detailed workup of T cells reactive with CT AA 32-40, we additionally studied a both wild-type peptide CT AA 32-40 peptide and a modified peptide in which AA 34 was modified from the wild-type cysteine residue to alanine. As detailed in Results, some CD8⁺ T cells do not efficiently react to 13 AA long peptides but show brisk recognition of shorter, internal peptides. Peptides were tested individually at 1 μg/mL final concentration by addition to Cos-7 cells (plated at 1.1 × 10⁴ cells/well prior to HLA transfection) 48 hours after HLA transfection. TILs (10⁵/well) were added 1–2 hours later, and 24–48-hour supernatants were tested for IFNγ by ELISA as indicated above. Alternatively, OLP were matrix-pooled into rows and columns of 9–10 peptides/pool and tested at 1 μg/mL final concentration each. Peptides at positive (mean ELISA OD₄₅₀ value > 0.2) pool intersection(s) were retested for confirmation. For some assays, aAPCs (Cos-7 cells plated at 1.1 × 10⁴ cells/well prior to HLA transfection) were peptide-pulsed at 10 μg/mL for 1 hour and PBS-washed before adding responder cells to reduce T-cell autopresentation. The HLA-peptide binding prediction algorithm was performed using netMHCpan 4.0 (http://www.cbs.dtu.dk/services/NetMHCpan/; ref. 36) at the Immune Epitope Database (IEBD; ref. 35).

To measure functional avidity, aAPCs were created by plating Cos-7 cells in 96-well plates and transiently transfecting with the relevant HLA cDNA for 2 days, as described above. Peptides were added in duplicate in serial 10-fold dilutions from 1 μg/mL to 1 pg/mL. Effector cells, either TCCs (5 × 10⁴/well) or bulk TILs (1 × 10⁵/well) were added for 2 days, after which supernatant IFNγ ELISA was done as described above. EC₅₀ values were estimated with Graphpad Prism V7.03. Nonlinear regression curve fitting used the dose–response stimulation log(agonist) versus normalized response algorithm.

For some subjects, T-Ag epitopes recognized by CD8⁺ T cells in TILs were mapped by intracellular cytokine secretion (ICS) flow cytometry exactly as reported previously (34). In brief, TILs were coincubated with autologous, carboxyfluorescein succinimidyl ester (CFSE, Invitrogen)-labeled PBMCs as APCs and OLP pools and were processed for ICS to detect TIL IFNγ accumulation after dump-gating of CFSE⁺ APCs as described previously (34). Positive pools were deconvoluted to individual peptides in follow-up assays.

HLA A*02:01-restricted CD8⁺ TCC w678.B11 is specific for LT AA 15-23 (20). TCC w830.5.1 was similarly cloned from subject w830 PBMC using tetramer sorting. TILs were stained with anti-CD8α-PE or APC (clone SK1, BioLegend) and then tetramer (below) and 7AAD (Thermo Fisher Scientific) for viability (27, 31). Gated single, live, CD8⁺, tetramer^bright cells were sorted (FACSAria II, Becton Dickinson) into 96-well plates and cloned as described previously (27, 28) using the FACSDiva software provided with the FACSAria II.

For IFNγ capture, aAPC were created by plating Cos-7 cells (5 × 10⁵) into a 12-well plate, and transfecting the next day with HLA B*35:03 cDNA and Fugene 6 (Promega) exactly as per the manufacturer's recommendation and similar to aAPC use in 96-well plates detailed above. After 48 hours, TILs (5 × 10⁶) and 1 μg/mL peptide were added in TCM for 18 hours. Cells were collected, incubated with the proprietary anti-IFNγ capture antibody reagent provided in the IFNγ Secretion Assay Cell Enrichment Kit (Miltenyi Biotec; 130-054-201) at 4°C for 20 minutes, and then returned to 37°C, 5% CO₂ for 45 minutes in 5 mL TCM. Cells were collected and stained with the proprietary phycoerythrin (PE)labeled anti-IFNγ clone provided in the kit, 7AAD, anti-CD3-APC (clone UCHT1, BioLegend), anti-CD4-APC-H7 (clone RPA-T4, Becton Dickinson), and anti-CD8-FITC (clone MHCD08014, Thermo Fisher Scientific). We gated on single, CD3⁺, CD4⁻, CD8⁺, IFNγ^bright cells and sorted them (FACSAria II) into 96-well plates for cloning. Analysis and gating used FACSDiva software. TCCs were screened for reactivity with 1 μg/mL peptide or media control and specific aAPCs, as detailed in Results. Screening assays performed at 12–16 days of TCC growth used approximately 20% of TCC cells per stimulus with HLA B*35:03–transfected aAPC and 1 mg/mL peptide CT AA 42-52, with IFNγ ELISA readout.

For TCCs, paired TRA and TRB complementarity determining region 3 (CDR3) regions were sequenced from cDNA. T cells (<500 cells/clone in <1 μL TCM) were added to TCL buffer (Qiagen). After RNA extraction (RNAClean XP RNA-SPRI, Beckman Coulter), cDNA was synthesized by SmartSeq2 5′-RACE (37) with primers, Oligo_dT, and LNA_TSO. cDNA was further PCR-amplified with primers for ISPCR. After cDNA purification (Agencourt Ampure XP beads, Beckman Coulter), cDNA within TRA and TRB, to include CDR3, were coamplified with library barcoding-compatible primers 4–7 (all primers listed in Supplementary Table S2). PCR products were barcoded using a custom Illumina barcode oligonucleotide matrix before quality control, normalization, pooling, and 2 × 300 basepair sequencing on an Illumina MiSeq (Illumina). TCR sequences were analyzed with the publicly available MiXCR program (ref. 38; https://mixcr.readthedocs.io/en/master/) and reference to the IMGT database (ref. 39; http://www.imgt.org). Epitopes and clonal CDR3 sequences were deposited at the Immune Epitope Database (IEDB; ref. 35) as ID 1000788. For ex vivo MCC tumor data, biopsy tissue (approximately 5 × 5 × 5 millimeter) was frozen in optimal cutting temperature (OCT) compound (Sakura). Scout sections stained with hematoxylin and eosin showed abundant MCC and infiltrating cells. Six adjacent 10-μm sections were submitted to Adaptive Biotechnology for genomic DNA extraction and Immunoseq TRB CDR3 sequencing and data processing, with data deposited at https://clients.adaptivebiotech.com/pub/jing-2020-cir.

Tetrameric complexes of HLA A*03:01, A*11:01, B*07:02, B*15:01, B*35:02, or B*37:01 and peptide conjugated to allophycocyanin (APC) or PE were made at Fred Hutchinson Cancer Research Center (Seattle, WA). Monomeric complexes of HLA B*44:02 and peptide (ImmunAware) were tetramerized with a 4:1 molar ratio of PE- or APC-labeled streptavidin (Thermo Fisher Scientific). TIL cells were stained with tetramer for 30 minutes at room temperature followed by addition of anti-CD8α-PE or -APC for 30 minutes at 4°C, washing, fixation, and analysis as specified above (20).

AA sequences of LT AA 1-330 and full-length ST from polyomaviruses detected in humans (Supplementary Table S4) were aligned using the tblastn routine within Geneious (Biomatters, v10.10.8) and default parameters.

MCC-derived MKL-1 and MS-1 cells are HLA A*03:01 and A*02:01-positive, respectively (14, 20, 40). Full HLA typing is provided above. The cells were a gift of Dr. Masa Shuda, University of Pittsburgh (Pittsburgh, PA) in 2012. The cells were authenticated by the ATCC Cell Line Authentication Service (https://www.atcc.org/en/Services/Testing_Services/Cell_Authentication_Testing_Service.aspx) with dates of last authentication for MKL-1 of May 24, 2019 and data of last authentication for MS-1 of December 1, 2016. Aliquots of cyropreserved (LN₂) cells were thawed a few passages prior to use rather than maintained in continuous culture. Both are maintained in RPMI1640 media with 1% penicillin/streptomycin, 2 mmol/L l-glutamine (all Thermo Fisher Scientific), and 10% FBS (Atlanta Biologicals). Cells were passaged in our lab approximately 20 times prior to use. Both were Mycoplasma-free on May 14, 2019 prior to use, tested by the Plasmotest Mycoplasma Detection Kit (Invivogen). Cells were treated 72 hours with rhIFNβ-1a (500 U/mL; PBL) and washed. HLA expression was detected via flow cytometry and PE-anti-HLA-A2 BB7.2 (Becton Dickinson) or PE-anti-HLA-A3 GAP_A3 (Thermo Fisher Scientific). MCC cells or control HLA-A*03:01–positive EBV-LCLs (2 × 10⁶ cells) were incubated with 10 μg/mL peptide MCPyV CT AA 32-40 or media for 1 hour followed by washes. HLA-A*03:01–positive EBV-LCL were cultured in-house exactly as described previously (28) from subject 1233. MCC or EBV-LCL cells (10⁵/well) and bulk tetramer-purified, TIL-origin MCPyV-specific CD8⁺ T cells (4 × 10⁵/well) were then coincubated 40 hours in TCM in triplicate in 96-well plates and secreted IFNγ was measured by ELISA.

Monocyte-derived DCs (MoDC) were cultured from cryopreserved, CD14-positive selected PBMCs (31). The LT 1-327 coding sequence from the NTC9385R-based plasmid, or enhanced GFP (eGFP) from eGFP-C1 (ref. 31; Takara), were PCR-amplified to add a T7 promoter (primers in Supplementary Table S2). PCR product was transcribed with a modified 5′ cap and 3′ polyadenylated (mMessage mMachine T7 Ultra, Thermo Fisher Scientific) and purified (MEGAclear, Thermo Fisher Scientific). DCs (day 8) were recovered, and 10⁶ were electroporated in 200 μL Opti-Mem (Thermo Fisher Scientific) with 3 μg mRNA using a Gene Pulser Xcell (Bio-Rad) at 300 V/500 microseconds. DCs and responder CD8⁺ T cells were seeded at 10⁵/well each into triplicate 96 well U-bottom in 200 μL TCM and supernatant IFNγ measured by ELISA at 24 hours.

The unpaired t test with Welch correction (Instat 3.0, Graphpad) was used to test IFNγ secretion from MCPyV-specific CD8⁺ T cells in response to MCPyV⁺ MCC cell line MKL-1 or positive control cells. A P value of <0.05 was considered significant.

We studied 12 persons with MCPyV⁺ MCC (Supplementary Table S1) using an aAPC approach. Each patient had T-Ag serum antibody, and subject 659 was tumor-positive for T-Ag protein by IHC. The median age was 66 (range, 21–79) and 50% were male. The MCC stage at biopsy ranged from small local to metastatic disease. Two subjects had an unknown site of primary disease. Two additional subjects' TILs (Supplementary Table S1) were studied using the published OLP ICS approach (34). Polyclonal PBMC-origin HLA B*35:02 tetramer-purified, MCPyV-specific CD8⁺ T cells from published, adoptive immunotherapy subject 2586-4 (24) were also tested.

We previously (20) reported that HLA A*02:01-restricted CD8⁺ TCCs recognize MCPyV antigen intracellularly processed from an LT AA 1-259–encoding plasmid. To detect potential CD8⁺ T-cell responses to C-terminal epitopes (2), we used an LT plasmid extended to AA 327 and closely matching the MCPyV consensus sequence (Supplementary Table S5). The AA 1-327 LT plasmid outperformed the AA 1-259 construct for stimulation of MCPyV-specific CD8⁺ TCCs recognizing an LT AA 15-23, an HLA A*02:01–restricted N-terminal epitope (Supplementary Fig. S1). The NTC LT plasmid and a similar ST plasmid (Supplementary Table S5) were used to probe TILs.

We interrogated TILs with an aAPC-expressing subject-specific HLA (27) and T-Ag. Bulk-expanded, polyclonal TILs frequently displayed IFNγ responses. For example, subject 1233 had responses for each of their HLA-A and -B alleles (Supplementary Fig. S1). For alleles A*03:01 and B*35:03, assays were positive for both LT and ST, whereas for B*37:01, only LT was antigenic.

Responses to HLA and T-antigen cotransfection, with negative responses to control aAPCs, were observed in eight additional MCPyV⁺ MCC TILs (Table 1; raw data, Supplementary Figs. S2 and S3). Overall, TILs from 9 of 12 (75%) of MCPyV⁺ MCCs recognized T-Ag in the context of one or more HLA class I alleles. In many subjects, multiple HLA alleles were active. Responses were detected for 8 HLA-B alleles (*08:01, *15:01, *18:01, *35:01, *35:03, *37:01, *44:02, and *57:01) and 4 HLA-A alleles (A*02:01, A*03:01, A*11:01, and A*24:02). We also probed TILs for HLA-C–restricted responses in most subjects, as published for alpha-herpes viruses (31), but did not observe reactivity (Supplementary Figs. S2 and S3).

To estimate the proportion of patients likely respond to CD8⁺ T-cell–directed immunotherapies, we compared the HLA class I alleles that restricted MCPyV T-Ag CD8⁺ T cells to the HLA types of 131 MCPyV⁺ MCC subjects. In addition to the alleles listed above, the restricting allele set included HLA A*23:01, previously shown (18) to restrict the same epitope presented by the related A*24:02 allele (34), HLA-B*35:02 as previously shown (24), and HLA-B*07:02, which we document below. For 59 subjects, with only two-digit HLA-A or -B typing available, we used the most prevalent AA-level allelic variant. For example, HLA-A*02:xx was assigned HLA*02:01. Analysis of HLA type data showed that, among 131 HLA-typed subject with MCPyV⁺ MCC, 127 of 131 (97%) had at least one active HLA allelic variant, with 126 of 131 (96%) and 109 of 131 (83%) expressing a relevant HLA-A or -B variant, respectively. Because MCPyV⁺ MCC can become resistant to CD8⁺ T-cell adoptive therapy specific for a single HLA class I–restricted epitope (24), we examined the percentage of patients with >1 potential CD8⁺ T-cell response and found that 108 of 131 (82%) expressed both at one or more relevant HLA-A and one or more relevant HLA-B alleles. MCPyV T-Ag, thus, has potential population coverage for CD8⁺ T-cell responses.

The aAPC cotransfection approach provided HLA restriction data and proof of antigen processing in initial positive assays, and in some cases, permitted preliminary epitope mapping. For example, the subject 1233 ST-positive, LT-negative results for HLA-B*37:01 (Supplementary Fig. S1) indicated that one or more epitopes were present in the C-terminal domain of ST. The reciprocal ST-negative, LT-positive pattern for subject 1107 and HLA-B*44:02 indicated epitope(s) in the C-terminal unique region of LT.

Tests of 13 AA-long T-Ag OLP explained some, but not all, aAPC results. For example (Fig. 1A), HLA-B*08:01–restricted reactivity with peptides CT AA 1-13 and CT AA 5-17 indicated a minimal epitope in the CT AA 5-13 overlap region. The reactive CT AA 29-41 peptide was similarly contained within both the ST and LT polypeptides, whereas the ST AA 133-145 was unique to ST. A modified OLP approach used matrix peptide pools. For example (Supplementary Fig. S4A), single row and column pools were reactive, each containing LT AA 97-109, within which the shorter AA 99-107 peptide was antigenic when tested singly. For each TIL-aAPC-13 AA peptide combination that was reactive, we followed up by studying internal peptides to determine shorter peptide epitopes (summarized in Supplementary Table S6). For some TILs, scans using 13 AA OLP were negative. Because there are constraints around trimming of termini and other steps for peptide-HLA loading that vary between HLA alleles and APCs (38, 39), peptides predicted to bind the relevant HLA allele(s) were tested. In an example for HLA-B*44:02 (Fig. 1B), TILs responded to LT AA 91-99, despite not having not reacted to the longer LT AA 89-101 peptide. Among the identified MCPyV CD8⁺ T-cell epitopes in this report, short peptides predicted to bind to relevant HLA alleles were required to obtain initial peptide reactivity for several epitopes (detailed in Supplementary Table S6).

HLA alleles cluster into supertypes based on peptide binding motifs (41). TCRs may also tolerate sequence variations in HLA allelic variants presenting the same peptide. We explored cross-recognition for the HLA-A3 supertype members A*03:01 and A*11:01. Subjects 1233 and 1243 recognized T-Ag in the contexts of endogenous A*03:01 and A*11:01, respectively. Subject 1233 TIL also recognized both ST and LT presented by non-self HLA-A*11:01 (Fig. 2). We tested 20 candidate peptides that were bioinformatically predicted, as described in Methods, to bind to HLA-A*03:01. Reactivity was detected for CT AA 31-40 and AA 32-40, considered to represent a single epitope, and for LT AA 95-103 and LT AA 293-302 (Fig. 2). The CT 31-40/32-40 peptides were also presented by non-self HLA-A*11:01, documenting promiscuity for both processed T-Ag and exogenous peptide.

Experiments with subject 1243 TILs, who express HLA-A*11:01 but not A*03:01, showed that each epitope discovered in the A*03:01 context was also antigenic in an HLA-A*11:01 MCC tumor. The data also confirmed CD8⁺ T-cell recognition of LT AA 95-103 by HLA-A*11:01, as has been reported (21). Even though TILs from subject 1243 recognized CT 31-40/32-40 using HLA*A11:01, reciprocal presentation by HLA-A*03:01 was not present (Fig. 2). In addition to the three epitopes also detected in subject 1233 in the HLA-A*03:01 setting, subject 1243 also had reactivity to LT AA 82-91, LT AA 261-270/262-270, and ST AA 125-134 and in each case, was restricted by HLA-A*11:01. Overall, this subject had the most diverse CD8⁺ T-cell TIL response to MCPyV T-Ag among those studied here, comprised of six HLA-A*11:01–restricted responses, and both HLA-A*02:01- and B*15:01–restricted responses (Table 1; Fig. 2).

In contrast to HLA-A*03:01/A*11:01, HLA-B*35:01 and -B*15:01 are not an HLA supertype (42). Nevertheless, TILs from subjects 1225 and 1243 (Supplementary Fig. S4) recognized the same peptide, LT AA 99-107, in the B*35:01 and B*15:01 contexts, respectively. Dual presentation of LT AA 99-107 by these two HLA alleles was confirmed with tetramers (Supplementary Fig. S5). Similarly, both HLA-B*37:01 and HLA-B*44:02 could presented LT AA 91-99. TILs from subject 735 recognized peptide LT AA 89-101 using HLA-B*37:01 (Supplementary Fig. S6). Subject 1056 (Supplementary Table S1) also recognized LT 91-99 (Supplementary Fig. S7) with the active HLA allele defined as B*37:01 using tetramers (Supplementary Fig. S5). Subject 1107 has CD8⁺ TILs that recognized LT presented by HLA-B*44:02 (Supplementary Fig. S2). Recognition was mapped to LT AA 91-99 using B*44:02 tetramers (Supplementary Fig. S5). The predicted high avidity of MCPyV T-Ag peptides for relevant HLA molecules included these instances of dual recognition within and outside of established HLA supertypes (Supplementary Table S6). Overall, we found it was common for CD8⁺ TILs to be able to use diverse HLA allelic variants to recognize short antigenic regions of MCPyV (Supplementary Table S6).

We probed the clonotypic complexity of subject 1233′s TILs that were reactive with CT AA 32-40 or LT AA 95-103. Tetrameric reagents of these peptides with HLA-A*03:01 stained discrete populations of CD8⁺ TILs (Supplementary Fig. S8A), which after sorting and expansion had reactivity to T-Ag and to peptide (Supplementary Fig. S8B). Functional testing of tetramer-sorted, single-cell origin clones revealed that for CT AA 32-40, 19 of 20 clones (95%) were MCPyV-reactive, as assessed by IFNγ responses to peptide-pulsed autologous EBV-LCL (Supplementary Fig. S8C). For the LT AA 95-103 epitope, 9 of 16 clones (56%) reacted to the peptide. TCR sequencing was successful for all 19 (100%) of the CT AA 32-40-specific TCCs and for 8 of 9 (89%) of the LT AA 95-103-specific TCCs. TCR sequencing showed five reactive TRB clonotypes specific for LT AA 32–40, among which, two had almost identical TRA and TRB CDR3 sequences (Table 2). In contrast, all sequenced LT 95-103–reactive clones were identical. To estimate the contribution of these MCPyV-specific TCR clonotypes to the tumor infiltrate, we performed quantitative TRB CDR3 sequencing from frozen sections of the tumor from which TILs had been cultured. All six of MCPyV-specific CD8⁺ T-cell clonotypes were detected in the ex vivo tumor tissue. Two CT AA 32-40 clonotypes were dominant, each comprising over 0.8% of the TCR sequences in the tumor sections and ranking 7th and 11th highest in overall abundance. The other clonotypes ranged from 0.02 to 0.19% in abundance (Table 2).

T-cell clonotypes specific for the same peptide/HLA can differ in tolerance for peptide or HLA variants. We dissected HLA-B*35–restricted responses because HLA-B*35 alleles differ for peptide binding, depending on HLA class I antigen processing and viral infection outcomes (43). Subject 1233′s bulk TILs recognized peptide CT AA 41-53 in the autologous B*35:03 context (Supplementary Fig. S9) but also stained with an HLA-B*35:02 tetramer containing CT AA 42-52 (Supplementary Fig. S10). Epitope mapping showed reactivity with CT AA 42–52 and 45–53 (Supplementary Fig. S9). Three discrete reactivity patterns for the minimal 9 AA peptide CT 45-53 were detected (Fig. 3) among relevant CD8⁺ clones (sorting pathway illustrated in Supplementary Fig. S10). Most clones, such as representative clone 1A1, recognized presentation via HLA-B*35:01 or B*35:02 in addition to the autologous B*35:03. However, others, such as clone 1C3, tolerated only B*35:02 or B*35:03, and one clone, 1C2, could use only the autologous B*35:03. (Fig. 3, top row). Each TCC recognized the 11 AA peptide CT 42-52 and also recognized epitopes processed from transfected MCPyV LT DNA, regardless of B*35 subtype (Fig. 3, middle and bottom rows). Overall, our results indicated that tumor-infiltrating MCPyV-specific CD8⁺ T clonotypes recovered by culture were also readily detected by molecular methods in fresh tumor tissue, and that depending on T-cell clonotype, the fine details of recognition by HLA allelic variants and peptides of varying length can vary.

The ability of CD8⁺ T cells to detect small amounts of peptide-HLA is thought to generally correlate with clinical efficacy. We estimated CD8⁺ T-cell functional avidity by using defined APCs, serial dilutions of peptide, and selected cloned and bulk CD8⁺ effector cells. For responses to MCPyV LT AA 32-40, we studied each a representative TCC for each clonotype detected by TCR sequencing in the context of HLA-A*03:01–expressing APCs (Table 2). These clones were derived from TIL from subject 1233 (Supplementary Table S1). Because the native peptide has a cysteine residue at AA 34 and cysteine can be unstable, we additionally studied an alanine-34 peptide homolog. Representative data for TCR clonotype 1 (Fig. 2) showed no difference in peptide EC₅₀ between peptides. For this epitope, EC₅₀ values for the native peptide ranged from 7.6 to 23.4 ng/mL (Table 2). For the LT AA 95-103 epitope and CD8⁺ TCC effectors, an EC₅₀ value of 1.9 ng/mL was detected (Fig. 2; Table 2). Using bulk TILs as effector cells, we also measured the functional avidity for three additional MCPyV epitopes ranging from 121 to 219 ng/mL (Supplementary Table S6). These functional avidity values were typical for CD8⁺ T cells.

We previously showed recognition of virus-positive MCC cell line MS-1 by HLA-A*02:01–restricted CD8⁺ TCCs (20). Tetramer-sorted, bulk-expanded, HLA-A*03:01–restricted CD8⁺ TILs reactive with CT AA 32–40 (Supplementary Fig. S8B) were unable to consistently recognize the HLA-A*03:01-positive, T-Ag–expressing MCC line MKL-1, even after IFN-β–induced upregulation of HLA expression (Supplementary Fig. S11). In the same experiments, CD8⁺ effector cells recognized peptide-pulsed EBV-LCLs used as APCs. We also noted that MKL-1 cells were only able to weakly present exogenous peptide to HLA-A*03:01–restricted CD8⁺ T cells, and only after IFNβ pretreatment (Supplementary Fig. S11). This indicated a suppressed interaction between these specific APCs and responder cells. Experiments to better characterize and overcome this immune evasion are underway. Because mRNA has been used as a cancer vaccine format (44), we induced LT expression in cultured monocyte-derived dendritic cells (moDC) by electroporation of LT mRNA. HLA-A*02:01–restricted MCC-specific TCC w678.B11 and CT 32-40-specific polyclonal CD8⁺ T cells restricted by HLA-A*03:01 each specifically recognized HLA-matched, LT-expressing dendritic cells (DC; Fig. 4).

Overall, we uncovered 16 CD8⁺ T-cell epitopes in T-Ag (Fig. 5). No examples of identify or near-identity with sequences in other polyomaviruses infecting humans (Supplementary Table S4) were noted. Epitope density near LT AA 100 prompted us to review the locations of CD8⁺ T-cell epitopes. We included additional MCPyV T-Ag epitopes not detailed above (peptide reactivity documented in Supplementary Figs. S12 and S13), as well as published epitopes (refs. 7, 20–22, 24, 34; summarized in Supplementary Table S6). Several epitopes were further validated with tetramers (Supplementary Figs. S5, S8, and S10). The CT AA 20-29 epitope, described previously (22) in the PBMCs of a subject with MCPyV⁺ MCC, was confirmed as antigenic within TILs from HLA-B*07:02–bearing subject 1179 using a HLA-B*07:02-peptide tetramer (Supplementary Fig. S5). Taking together published and MCPyV CD8⁺ T-cell epitopes identified here, we observed spatial clustering (Fig. 5), particularly in the MCPyV LT AA 70-110 domain. The biological underpinnings of T-cell epitope clustering are presently unknown. We previously showed that the LT AA 212-216 region with the retinoblastoma protein-binding LxCxE (45) domain contains multiple CD4⁺ T-cell epitopes and reported AA changes that can reduce oncogenic potential without changing antigenicity (46). This region is not CD8⁺ T-cell epitope rich, suggesting that changes in this domain would not impair vaccine CD8⁺ T-cell immunogenicity.

In this report, we documented that most MCPyV⁺ MCC tumors are infiltrated with CD8⁺ T cells that recognize the driver viral T-Ag oncoproteins. Diverse HLA restriction indicated that most patients could benefit from optimized immunotherapy to boost endogenous, or provide exogenous, MCPyV-specific T cells. The small and well-defined CD8 antigenic space in MCPyV⁺ T-Ag now provides an opportunity to conduct mechanistic studies of licensed and candidate immunotherapies, with the ability to track antigen-specific CD8⁺ T cells in a large proportion of patients.

Despite negative results of some phase III trials and limited licensure (47), there is great hope that cancer vaccines can become more successful. Microbial proteins in pathogen-associated malignancies may have an advantage over native tumor antigens due to reduced self-tolerance. MCPyV⁺ MCC could be appropriate to vaccine approaches due to restricted antigenic complexity and the ease of tracking antigen-specific responses, as well as the documented presence of CD8⁺ T-cell epitopes covering the vast majority of the population.

The TILs we studied here have not completely rejected tumors, presumably due to tumor microenvironment, immunoregulatory cells, tumor cell–intrinsic mechanisms, or T-cell exhaustion. We were not successful in consistently observing recognition of MCPyV⁺ MCC cell lines with culture-expanded TILs, even after boosting HLA class I expression. This implied that continuing studies of the immune-suppressive mechanisms in MCC tumors are vital. It is also possible that the expanded TILs we studied were subtly dysfunctional for recognition of tumor cells, despite reactivity to other APCs. It is likely that vaccination or adoptive cell therapy that bolsters T-Ag–specific T cells will need to be combined with measures that modify immune evasion in MCC tumors to obtain significant antitumor effects in the majority of persons. Licensed drugs that inhibit the PD-1 pathway, as well as local therapy with cytokines, oncolytic viruses, and innate immune agonists, each have activity in MCC (5–8, 10), providing candidate approaches to combination therapy.

Many projects are mining the CD8⁺ TIL repertoire to identify tumor antigens, principally mutated neoantigens, with the goal of creating personalized cancer vaccines or adoptively transferred TCR products (48). The technical challenges imposed by large mutant peptidomes and the use of cloned T cells or TCRs re-created from TIL TCR sequencing are slowing these efforts. In contrast to other immune-responsive tumors, the mutational burden in MCPyV⁺ MCC is extraordinarily low. The potential epitope space in MCPyV T-Ag is small, and T-Ag sequences are relatively invariant, such that vaccine construction is simplified. Tumor cells are known to require continuing expression of MCPyV T-Ag for cell proliferation (2), limiting the potential for immune escape due to T-Ag loss. We have shown that MCPyV⁺ MCC can adapt to CD8⁺ T-cell pressure by downmodulating expression of either HLA-A or -B via epigenetic mechanisms (24). Vaccine-boosting of broad, HLA A-, and -B–restricted CD8⁺ T-cell responses, or infusion of T cells restricted by multiple HLA alleles, could minimize escape by selective HLA allelic downregulation. The tools provided in this report suggest that a T-Ag–based MCC vaccine could be broadly immunogenic and that multi-epitope adoptive cell therapy should also be possible in many persons with diverse HLA haplotypes.

T-Ag DNA vaccination of mice can prevent or treat tumors caused by cells that express MCPyV T-Ag (11, 12). Using human cells, Gerer and colleagues showed that repeated stimulation of PBMCs with LT mRNA–transduced moDCs can activate autologous LT-specific T cells from patients with MCPyV⁺ MCC and healthy persons (13). These prototype vaccines explored fusions of T-Ag with host molecules designed to improve antigen presentation. In contrast, we used plasmid DNA or mRNA constructs without an MHC-targeting moiety to achieve stimulation of MCPyV-specific CD8⁺ T cells. Future work can study the relative efficiency of enhancements to promote antigen processing. Because the priming of new clonotypes may be critical (49), the presence of well-documented, immunodominant T-Ag CD8 and CD4 epitopes in mice (11, 12, 50) can be used to compare vaccine priming strategies prior to advancement to humans.

The critical domains for MCPyV ST and truncated LT promotion of cell proliferation are incompletely understood (2). T-Ag–encoding nucleic acids within a vaccine will likely need to be genetically detoxified, analogous to the HPV-E6 and -E7 candidates in clinical-stage vaccines (51). It is likely CD4⁺ T cells participate in effective antitumor T-cell response (52). Our preliminary map of T-Ag CD8⁺ T-cell epitope–rich regions, together with our previous CD4⁺ T-cell work (46) and ongoing mapping of T-Ag oncogenic regions will inform safety modifications to T-Ag in vaccine candidates and also may enable the design of polyepitope immunogens.

Our aAPC approach to CD8⁺ T-cell detection has both strengths and limitations. It is sensitive, does not require PBMCs, is usable on small biopsies, and gives definitive HLA restriction data as soon as positive reactivity is detected. Relatively little MCPyV strain variation in T-Ag polypeptides have been detected, such that universal T-Ag reagents can be used. Although we have not yet detected any examples, epitopes created by posttranslational modification may also be present in transfected aAPCs. Caveats include the possibility that CD8⁺ T-cell responses to C-terminal nonsense polypeptides that could occur distal to LT frameshift mutations would not be missed with our LT plasmid. If TILs are extremely dysfunctional, it is possible they will fail to expand in vitro. Finally, use of IFNγ for detection of T-cell activation could miss some T cells, a concern somewhat ameliorated by the consensus that even dysfunctional CD8⁺ T cells usually retain IFNγ responses (53).

The current report extends previous knowledge of T-cell responses to MCPyV T-Ag. Prior reports have used both TILs and blood from healthy persons and patients with MCC, and in some cases, not differentiating between memory responses and the naïve T-cell repertoire (21–23). Our work with TILs confirms some of the T-Ag epitopes found with high-throughput multiplexed PBMC tetramer methods. Our findings complement a study by Samimi and colleagues of TILs from 18 persons with MCPyV⁺ MCC (21). Using bulk CD8⁺ TIL effectors and a similar aAPC cotransfection approach, they documented unique LT-specific responses in 5 of 18 (28%) persons with MCPyV⁺ MCC, a lower proportion than in our report. The HLA-B*18:01/LT 77-85 and HLA-A*11:01/LT 95-103 epitopes reported by Samimi and colleagues are confirmed herein. In contrast to our work, Samimi and colleagues did not observe CD8⁺ T-cell responses to the MCPyV CT or ST domains. Our studies concur that LT, especially the AA 70–110 region, appears to be immunodominant for CD8⁺ T-cell responses. The reasons for contrasting results could relate to differences in the T-cell activation readout, patient populations, plasmid constructs, or TIL culture conditions. There were also differences in methods for HLA overexpression in MCC cell lines, with our difficulty in consistently achieving CD8⁺ T-cell recognition of MCC cells highlighting the need for further research.

In summary, most MCPyV⁺ MCC tumors were infiltrated by readily expandable CD8⁺ T cells that recognized viral peptides in the context of HLA-A or -B molecules. The response that is often polyepitope–specific and uses both HLA-A and -B. Studies to correlate virus-specific CD8⁺ T cells in TILs, or blood, and outcome during and after licensed and experimental immunotherapy can use the expanded toolkit described herein. The clustering of CD8⁺ T-cell immunogenic regions in specific portions of MCPyV LT suggest strategies for vaccine design to maintain high population coverage while reducing the potential for oncogenic toxicity. In particular, the region around AA 70–110 of LT is particularly rich in CD8⁺ T-cell epitopes. Despite the immunogenicity of MCPyV T-Ag and the presence of relevant virus-specific TILs, tumors persist, and growth. The findings in this report indicated that passive or active immunotherapy, enhanced to overcome immune evasion, has the potential to help the broad spectrum of patients with MCPyV⁺ MCC.

C.D. Church has ownership interest (including patents) in University of Washington. K.G Paulson reports receiving commercial research grants from SITC-Merck Fellowship and Bluebird Biosciences. S. Bhatia is an advisory board member for EMD Serono, Sanofi-Genzyme, and Bristol-Myers Squibb and reports receiving commercial research grants from EMD Serono, Bristol-Myers Squibb, Merck, Immune Design, NantKwest, Exicure, Novartis, and OncoSec. P. Nghiem is a consultant for EMD Serono; reports receiving commercial research grants from EMD Serono and Bristol-Myers Squibb; and has ownership onterest (including patents) in intellectual property owned by the University of Washington, T cells for Merkel cell carcinoma. D.M. Koelle reports receiving a commercial research grant from Immunomics Therapeutics and is a co-inventor on patent applications by Fred Hutchinson Cancer Research Center and the University of Washington on Merkel cell polyomavirus-specific T cells, including therapies using T-cell receptors, and vaccines. No potential conflicts of interest were disclosed by the other authors.

Conception and design: L. Jing, A. Colunga, D.M. Koelle

Development of methodology: L. Jing, M.M. Cook, A.L. Greninger, D.M. Koelle

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): M. Ott, C.D. Church, R.M. Kulikauskas, D. Ibrani, J.G. Iyer, O.K. Afanasiev, A. Colunga, M.M. Cook, H. Xie, K.G. Paulson, S. Bhatia, P. Nghiem

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): L. Jing, M. Ott, C.D. Church, A. Colunga, M.M. Cook, A.G. Chapuis, D.M. Koelle

Writing, review, and/or revision of the manuscript: L. Jing, M. Ott, C.D. Church, R.M. Kulikauskas, J.G. Iyer, M.M. Cook, A.L. Greninger, K.G. Paulson, A.G. Chapuis, S. Bhatia, P. Nghiem, D.M. Koelle

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): L. Jing, M. Ott, C.D. Church, R.M. Kulikauskas, D. Ibrani, P. Nghiem

Study supervision: L. Jing, D.M. Koelle

The authors gratefully acknowledge assistance from Christopher McClurkan, Jianhong Cao, PhD, at the Fred Hutchinson Cancer Research Center Immune Monitoring Shared Resource, Anette Stryhn, PhD, and Soren Buus, PhD, at ImmunAware for HLA-peptide reagents, and Sandra G. Porter, PhD, and Todd M. Smith, PhD, at Digital World Biology for graphic arts assistance. This work was supported by the Immune Monitoring Shared Resource of the Fred Hutch/University of Washington Cancer Consortium (NIH/NCI Cancer Center Support Grant P30 CA015704), NIH grants (P01-CA225517, K24-CA139052, R01-CA162522, and R01-CA176841; to P. Nghiem), Kelsey Dickson Team Science Courage Research Team Award, Prostate Cancer Foundation Award (#15CHAS04, to P. Nghiem), NIH Center for AIDS Research (CFAR) NIH grant AI027757, the David & Rosalind Bloom Endowment for MCC Research (to P. Nghiem), the UW MCC Patient Gift Fund (to P. Nghiem), and Friends and Family of Nancy Haeseker (to P. Nghiem).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.